BitSage Network - GPU Worker Node

╔═══════════════════════════════════════════════════════════════════════════════╗
║ ║
║ ██████╗ ██╗████████╗███████╗ █████╗ ██████╗ ███████╗ ║
║ ██╔══██╗██║╚══██╔══╝██╔════╝██╔══██╗██╔════╝ ██╔════╝ ║
║ ██████╔╝██║ ██║ ███████╗███████║██║ ███╗█████╗ ║
║ ██╔══██╗██║ ██║ ╚════██║██╔══██║██║ ██║██╔══╝ ║
║ ██████╔╝██║ ██║ ███████║██║ ██║╚██████╔╝███████╗ ║
║ ╚═════╝ ╚═╝ ╚═╝ ╚══════╝╚═╝ ╚═╝ ╚═════╝ ╚══════╝ ║
║ ║
║ The Economic Heart of Decentralized Compute ║
║ ║
╠═══════════════════════════════════════════════════════════════════════════════╣
║ ║
║ ██████╗ ██████╗ ███████╗██╗ ██╗ ██╗███████╗██╗ ██╗ ║
║ ██╔═══██╗██╔══██╗██╔════╝██║ ╚██╗ ██╔╝██╔════╝██║ ██╔╝ ║
║ ██║ ██║██████╔╝█████╗ ██║ ╚████╔╝ ███████╗█████╔╝ ║
║ ██║ ██║██╔══██╗██╔══╝ ██║ ╚██╔╝ ╚════██║██╔═██╗ ║
║ ╚██████╔╝██████╔╝███████╗███████╗██║ ███████║██║ ██╗ ║
║ ╚═════╝ ╚═════╝ ╚══════╝╚══════╝╚═╝ ╚══════╝╚═╝ ╚═╝ ║
║ ║
║ Verifiable GPU Compute • ZK Proofs • TEE Attestation ║
║ ║
╚═══════════════════════════════════════════════════════════════════════════════╝

BitSage Network - GPU Worker Node

High-performance Rust node for the BitSage Network, featuring Obelysk Protocol integration with GPU-accelerated zero-knowledge proofs.

One-Command Deployment

Option 1: Docker (Recommended)

Start earning SAGE tokens with a single command:

# GPU Worker (requires nvidia-docker)
docker compose -f docker-compose.worker.yml up
# CPU-only (for testing)
docker compose -f docker-compose.worker.yml --profile cpu up

This automatically:

Generates a wallet (account abstraction)
Claims tokens from faucet (testnet)
Registers with the coordinator
Starts processing jobs
Earns SAGE tokens

Option 2: Native Installation

curl -sSL https://raw.githubusercontent.com/Bitsage-Network/rust-node/main/scripts/install.sh | bash

This interactive wizard will:

Detect your GPU and system configuration
Install all dependencies (Rust, CUDA drivers)
Download and build the sage-worker
Generate wallet and encryption keys
Register with the network
Start earning immediately

Manual Installation

Requirements

Component	Minimum	Recommended
GPU	NVIDIA RTX 3090 (24GB)	NVIDIA H100 (80GB)
CUDA	12.0+	12.4+
RAM	32 GB	64 GB
Storage	100 GB SSD	500 GB NVMe
Network	100 Mbps	1 Gbps
OS	Ubuntu 22.04	Ubuntu 22.04/24.04

Build from Source

# Clone the repository
git clone https://github.com/Bitsage-Network/rust-node
cd bitsage-network/rust-node
# Build the worker
cargo build --release --bin sage-worker
# Run the setup wizard
./target/release/sage-worker setup --network sepolia
# Start the worker
./target/release/sage-worker start

What the Setup Wizard Does

Detects your GPU(s) automatically
Generates a Starknet wallet (or imports existing)
Generates ElGamal encryption keys for privacy
Registers with the BitSage coordinator
Saves configuration to ~/.bitsage/worker.toml

Start Earning

# Start your worker
sage-worker start
# Check status
sage-worker status
# View earnings
sage-worker info

Staking Tiers

Stake SAGE tokens to unlock higher-paying jobs:

Tier	Stake Required	Job Priority	Max Concurrent Jobs
Consumer	1,000 SAGE	Standard	4
Workstation	2,500 SAGE	Priority	8
DataCenter	5,000 SAGE	High Priority	16
Enterprise	10,000 SAGE	Premium	32
Frontier	25,000 SAGE	Maximum	Unlimited

# Stake tokens
sage-worker stake --amount 5000
# Check stake status
sage-worker info
# Claim rewards
sage-worker claim

Performance Benchmarks

Latest Results (February 3, 2026)

Hardware: NVIDIA H100 PCIe 80GB Network: Starknet Sepolia Events per TX: 12 (full fee distribution cascade)

Proof Generation Performance

Workload	Trace Steps	GPU Time	CPU Time	Speedup	Gas Cost
ML Inference	132	21ms	18ms	1.0x	0.307 STRK
1K Steps	1,024	24ms	20ms	0.8x	0.307 STRK
64K Steps	65,536	159ms	164ms	1.0x	0.380 STRK
256K Steps	262,144	335ms	352ms	1.1x	0.404 STRK
1M Steps	1,048,576	1,107ms	1,125ms	1.0x	0.429 STRK

On-Chain Verification Stats

Metric	Value
Events per TX	12
Internal Calls	10+
Success Rate	100%
Calldata Size	173-333 felts
FRI Layers	8-20

Verified Transactions (Live on Starknet Sepolia)

ML_GPU: https://sepolia.voyager.online/tx/0x068545dbe5b18a52328b0c0b74a661c6f0f7f689d4847247b055bd217a46cf53
ML_CPU: https://sepolia.voyager.online/tx/0x051ee2466af84d94b439fae15bcb1662317a4a7116ee3e7ccb3a3f07ae731eac
GPU_1K: https://sepolia.voyager.online/tx/0x03962dcd9b61dbcd7e5f24fab76132ad29ba4c6ba6e3b667b7f78055ee876e72
GPU_64K: https://sepolia.voyager.online/tx/0x03cc26baf34abbed4c753ce60e53854d8728723a73acc3f7fa9f687fc6f9bfb1
GPU_256K: https://sepolia.voyager.online/tx/0x0384d3daa5f08e083115c228b91d19a2a79d3d73117eb57f666f9ec8b3574607
GPU_1M: https://sepolia.voyager.online/tx/0x05d0ae5280523e1ec31802a8aa7ffec28eea943c498d7b1694a495087557eec9
CPU_1M: https://sepolia.voyager.online/tx/0x03494f9bd7eb9e5a1b323b12e0478d12876d8c943b9b92035b61d824ecd8a2fe

Multi-GPU Scaling (4x H100)

Metric	Value
Throughput	1,237 proofs/sec
Per-proof time	0.81ms
Scaling efficiency	193% (super-linear!)
Hourly capacity	4.45 million proofs
Daily capacity	107 million proofs

GPU Comparison

GPU	Speedup	Proofs/sec	Monthly Earnings*
RTX 3090	15-40x	~50	~$200-400
RTX 4090	25-60x	~80	~$350-600
A100 80GB	45-130x	127	~$500-900
H100 80GB	55-174x	150	~$600-1,100
4x H100	55-174x	1,237	~$2,500-4,500

*Estimated based on network demand and staking tier

Architecture

bitsage-network/
├── rust-node/
│ ├── src/
│ │ ├── bin/
│ │ │ ├── sage_worker.rs # GPU Worker CLI (run this!)
│ │ │ ├── proof_cli.rs # Proof generation CLI
│ │ │ └── unified_coordinator.rs
│ │ ├── obelysk/ # Obelysk Protocol
│ │ │ ├── prover.rs # ZK proof generation
│ │ │ ├── vm.rs # Obelysk Virtual Machine
│ │ │ ├── gpu/ # GPU acceleration
│ │ │ └── stwo_adapter.rs # Stwo GPU integration
│ │ ├── coordinator/ # Job coordination
│ │ ├── network/ # P2P networking
│ │ ├── blockchain/ # Starknet integration
│ │ └── compute/ # Job execution
│ └── libs/stwo/ # GPU-accelerated Stwo prover

How It Works

┌─────────────────────────────────────────────────────────────────┐
│ BitSage Network Architecture │
├─────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────────────┐ │
│ │ Your GPU │ │
│ │ sage-worker │◄────── You run this │
│ └────────┬─────────┘ │
│ │ │
│ │ HTTPS (polls for jobs) │
│ ▼ │
│ ┌──────────────────┐ │
│ │ Coordinator │◄────── Managed by BitSage │
│ │ (BitSage) │ │
│ └────────┬─────────┘ │
│ │ │
│ │ On-chain verification │
│ ▼ │
│ ┌──────────────────┐ │
│ │ Starknet │◄────── Smart contracts │
│ │ Blockchain │ (ProverStaking, JobManager) │
│ └──────────────────┘ │
│ │
└─────────────────────────────────────────────────────────────────┘

Obelysk Proof Pipeline

┌─────────────────────────────────────────────────────────────────┐
│ Obelysk Proof Pipeline │
├─────────────────────────────────────────────────────────────────┤
│ │
│ 1. Client submits encrypted workload │
│ │ │
│ ▼ │
│ 2. Data uploaded to GPU (stays in TEE) │
│ │ │
│ ▼ │
│ 3. GPU computes: FFT → FRI → Merkle │
│ (Data NEVER leaves GPU - 174x faster!) │
│ │ │
│ ▼ │
│ 4. 32-byte proof/attestation returned │
│ │
└─────────────────────────────────────────────────────────────────┘

Obelysk Protocol: Understanding the Proving System

Overview

BitSage uses a two-layer proving architecture:

┌─────────────────────────────────────────────────────────────────┐
│ YOUR APPLICATION │
│ (ML inference, data processing, privacy) │
└───────────────────────────┬─────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ OBELYSK PROTOCOL │
│ High-level proving API — what YOU interact with │
│ │
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ ObelyskVM │ │ Prover │ │ Starknet │ │
│ │ (M31 field) │ │ (circuits) │ │ (verify) │ │
│ └──────────────┘ └──────────────┘ └──────────────┘ │
└───────────────────────────┬─────────────────────────────────────┘
│ stwo_adapter.rs
▼
┌─────────────────────────────────────────────────────────────────┐
│ STWO PROVER (libs/stwo/) │
│ Low-level Circle STARK cryptography — handled internally │
│ │
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ Circle FFT │ │ FRI │ │ Merkle │ │
│ │ (GPU) │ │ (folding) │ │ (commit) │ │
│ └──────────────┘ └──────────────┘ └──────────────┘ │
└─────────────────────────────────────────────────────────────────┘

Key Points:

Obelysk = BitSage's protocol layer (high-level, user-facing)
Stwo = StarkWare's Circle STARK prover (low-level, internal)
You interact with Obelysk APIs — Stwo is handled automatically
The stwo_adapter.rs file bridges the two layers

Why Two Layers?

Layer	Purpose	You Touch It?
Obelysk	Application logic: ML, privacy, payments	✅ Yes
Stwo	Cryptographic proving (Circle STARKs)	❌ No (internal)

Benefits:

Obelysk provides simple APIs for verifiable computation
Stwo provides 10-174x GPU speedup via Circle FFTs
Mersenne-31 field (2³¹ - 1) enables native 32-bit operations
Clean separation means you focus on business logic, not cryptography

Developer Guide: Generating Proofs

Quick Start: Generate Your First Proof

# Build with GPU support
cargo build --release --features cuda
# Generate a demo proof (proves ML inference)
./target/release/bitsage-proof demo --batch-size 100
# Output:# ✓ Generated proof for 100 ML inferences# ✓ Proof size: 48.2 KB# ✓ Proving time: 127ms (GPU)# ✓ Proof hash: 0x7a3f...

CLI Proof Generation

# Generate proof for a workload
bitsage-proof generate \
--workload ml-inference \
--batch-size 1000 \
--output proof.json
# Verify locally (fast, no gas)
bitsage-proof verify --proof proof.json
# Submit to Starknet (costs gas, permanent)
bitsage-proof submit \
--proof proof.json \
--network sepolia
# Check on-chain status
bitsage-proof status --tx-hash 0x...

Rust API: Programmatic Proof Generation

use bitsage_node::obelysk::{ObelyskVM,ObelyskProver,ProverConfig,Instruction,OpCode,M31,};fnmain() -> anyhow::Result<()>{// 1. Create and execute your computationletmut vm = ObelyskVM::new();// Simple example: compute 2 * 3 + 5 = 11
vm.load_program(vec![Instruction::new(OpCode::LoadImm,0,2,0),// r0 = 2Instruction::new(OpCode::LoadImm,1,3,0),// r1 = 3Instruction::new(OpCode::Mul,2,0,1),// r2 = r0 * r1 = 6Instruction::new(OpCode::LoadImm,3,5,0),// r3 = 5Instruction::new(OpCode::Add,4,2,3),// r4 = r2 + r3 = 11]);let trace = vm.execute()?;println!("Result: {:?}", trace.final_state());// 2. Generate ZK prooflet prover = ObelyskProver::new(ProverConfig::default());let proof = prover.prove_execution(&trace)?;println!("Proof size: {} bytes", proof.size());println!("Proof hash: 0x{}", hex::encode(&proof.commitment()));// 3. Verify locallyassert!(prover.verify(&proof)?);println!("✓ Proof verified!");Ok(())}

GPU-Accelerated Proving

use bitsage_node::obelysk::{prewarm_gpu, stwo_adapter};fnmain() -> anyhow::Result<()>{// Pre-warm GPU (compile CUDA kernels)prewarm_gpu()?;// Generate trace (same as before)let trace = generate_ml_trace()?;// Prove with GPU acceleration (50-174x faster)let proof = stwo_adapter::prove_with_stwo_gpu(&trace)?;println!("GPU proving time: {:?}", proof.metrics.total_time);Ok(())}

Submitting Proofs to Starknet

use bitsage_node::obelysk::{StarknetClient,ProofSerializer};asyncfnsubmit_proof(proof:StarkProof) -> anyhow::Result<String>{// Connect to Starknetlet client = StarknetClient::new("https://starknet-sepolia.public.blastapi.io","0x...",// Your private key).await?;// Serialize proof to Cairo formatlet cairo_proof = ProofSerializer::serialize(&proof)?;// Submit to verifier contractlet tx_hash = client.submit_proof(
cairo_proof,"0x..."// Verifier contract address).await?;println!("Submitted! TX: {}", tx_hash);Ok(tx_hash)}

Available Workload Types

Workload	Description	Use Case
`ml-inference`	Neural network inference proof	Verify ML model ran correctly
`data-transform`	ETL pipeline proof	Verify data processing
`privacy-transfer`	Private payment proof	Confidential transactions
`batch-verify`	Aggregate multiple proofs	Gas-efficient verification
`custom`	User-defined computation	Any ObelyskVM program

Module Reference

Core Obelysk Modules

Module	Purpose	Key Types
`obelysk::vm`	Virtual machine execution	`ObelyskVM`, `Instruction`, `OpCode`
`obelysk::prover`	Proof generation	`ObelyskProver`, `ProverConfig`, `StarkProof`
`obelysk::field`	Mersenne-31 arithmetic	`M31`
`obelysk::circuit`	Circuit building	`Circuit`, `CircuitBuilder`
`obelysk::stwo_adapter`	Stwo integration	`prove_with_stwo`, `prewarm_gpu`

GPU Modules

Module	Purpose	Key Types
`obelysk::gpu`	GPU acceleration	`GpuProver`, `GpuConfig`
`obelysk::gpu::fft`	Circle FFT on GPU	(internal)
`obelysk::gpu::memory_pool`	GPU memory management	(internal)

Starknet Modules

Module	Purpose	Key Types
`obelysk::starknet`	On-chain verification	`StarknetClient`, `VerifierContract`
`obelysk::proof_packer`	Proof serialization	`PackedProof`, `pack_proof`

Privacy Modules

Module	Purpose	Key Types
`obelysk::elgamal`	ElGamal encryption	`encrypt`, `decrypt`, `KeyPair`
`obelysk::privacy_client`	Private transfers	`PrivateAccount`, `PrivatePayment`
`obelysk::fhe`	Homomorphic encryption	`FheEncryptor`, `EncryptedValue`

CLI Reference

sage-worker Commands

# Initial setup
sage-worker setup --network mainnet
# Worker management
sage-worker start # Start worker
sage-worker stop # Stop worker
sage-worker status # Check status
sage-worker logs # View logs# Staking
sage-worker stake --amount 5000 # Stake SAGE tokens
sage-worker unstake --amount 1000 # Unstake tokens
sage-worker claim # Claim rewards# Information
sage-worker info # Show worker info + earnings
sage-worker export# Export wallet (backup)

bitsage-proof Commands

# Generate proofs manually
bitsage-proof generate \
--batch-size 1000 \
--security-bits 128 \
--output proof.json
# Generate TEE attestation
bitsage-proof attest \
--proof proof.json \
--output quote.bin
# Submit to Starknet
bitsage-proof submit \
--proof proof.json \
--quote quote.bin \
--network mainnet

Configuration

Worker Configuration (~/.bitsage/worker.toml)

worker_id = "worker-abc12345"network = "mainnet"coordinator_url = "https://coordinator.bitsage.network"starknet_rpc = "https://starknet-mainnet.public.blastapi.io"dashboard_url = "https://dashboard.bitsage.network"
[wallet]
address = "0x..."private_key_path = "~/.bitsage/keys/starknet.key"elgamal_key_path = "~/.bitsage/keys/elgamal.key"
[gpu]
detected = truecount = 4model = "NVIDIA H100 80GB HBM3"memory_gb = 80compute_capability = "9.0"tee_supported = true
[settings]
poll_interval_secs = 5heartbeat_interval_secs = 30max_concurrent_jobs = 16auto_claim_rewards = true

Environment Variables

# Override config fileexport BITSAGE_NETWORK=mainnet
export BITSAGE_COORDINATOR_URL=https://coordinator.bitsage.network
export STARKNET_PRIVATE_KEY=0x...
# GPU settingsexport CUDA_VISIBLE_DEVICES=0,1,2,3 # For multi-GPU

Building from Source

Standard Build (CPU only)

cargo build --release

GPU Acceleration (NVIDIA CUDA)

# Requires CUDA Toolkit 12.x
cargo build --release --features cuda

Multi-GPU Support

cargo build --release --features cuda,multi-gpu

All Features

cargo build --release --features cuda,gpu-metrics,redis-cache,tee

Feature Flags

Feature	Description
`cuda`	NVIDIA CUDA GPU acceleration
`gpu-metrics`	NVIDIA NVML metrics monitoring
`rocm`	AMD ROCm support (future)
`tee`	Trusted Execution Environment support
`redis-cache`	Redis caching for dashboard
`fhe`	Fully Homomorphic Encryption

Docker Deployment

Using Docker

# Pull the image
docker pull ghcr.io/ciro-ai-labs/sage-worker:latest
# Run with GPU support
docker run --gpus all \
-v ~/.bitsage:/root/.bitsage \
-e STARKNET_PRIVATE_KEY=$STARKNET_PRIVATE_KEY \
ghcr.io/ciro-ai-labs/sage-worker:latest \
start

Docker Compose

version: '3.8'services:
sage-worker:
image: ghcr.io/ciro-ai-labs/sage-worker:latestdeploy:
resources:
reservations:
devices:
- driver: nvidiacount: allcapabilities: [gpu]volumes:
- ~/.bitsage:/root/.bitsageenvironment:
- BITSAGE_NETWORK=mainnet
- STARKNET_PRIVATE_KEY=${STARKNET_PRIVATE_KEY}restart: unless-stopped

Cloud Deployment

AWS (EC2 P4d/P5)

# Launch H100 instance
aws ec2 run-instances \
--instance-type p5.48xlarge \
--image-id ami-0abcdef1234567890 \
--key-name your-key
# SSH and install
ssh -i your-key.pem ubuntu@<instance-ip>
curl -sSL https://get.bitsage.network/sage-worker | bash
sage-worker setup --network mainnet
sage-worker start

Lambda Labs

# Launch from Lambda dashboard or CLI
lambda instance create --type gpu_8x_h100 --region us-west-1
# SSH and install
ssh ubuntu@<instance-ip>
curl -sSL https://get.bitsage.network/sage-worker | bash
sage-worker setup --network mainnet
sage-worker start

Brev.dev (Kubernetes)

# Deploy via Brev
brev deploy --gpu h100 --count 4
# Worker starts automatically via init container

Monitoring

Built-in Dashboard

Access your worker status at: https://dashboard.bitsage.network

Prometheus Metrics

Metrics are exposed at http://localhost:9090/metrics:

# HELP sage_proofs_generated_total Total proofs generated
sage_proofs_generated_total 12345
# HELP sage_gpu_utilization_percent Current GPU utilization
sage_gpu_utilization_percent{gpu="0"} 87.5
# HELP sage_earnings_sage Total SAGE earned
sage_earnings_sage 1234.56

Grafana Integration

Import the BitSage Grafana dashboard:

# Dashboard ID: 12345

Troubleshooting

Common Issues

GPU not detected

# Check NVIDIA driver
nvidia-smi
# Check CUDA
nvcc --version
# Reinstall CUDA toolkit
sudo apt install cuda-toolkit-12-4

Connection refused

# Check coordinator URL
curl -s https://coordinator.bitsage.network/health
# Check firewall
sudo ufw allow 443/tcp

Wallet issues

# Export and backup wallet
sage-worker export> wallet-backup.json
# Check balance on Starknet
starkli balance <your-address> --network mainnet

Low earnings

# Increase stake for priority jobs
sage-worker stake --amount 5000
# Check job history
sage-worker info --verbose

Logs

# View worker logs
sage-worker logs
# View system logs
journalctl -u sage-worker -f
# Debug mode
RUST_LOG=debug sage-worker start

Security

Key Management

Private keys are stored encrypted in ~/.bitsage/keys/
Never share your private key or export file
Use a hardware wallet for large stakes (coming soon)

TEE Attestation

Workers with TEE support (H100, TDX, SEV) generate cryptographic attestation:

# Generate attestation quote
bitsage-proof attest --output quote.bin
# Verify attestation
bitsage-proof verify-quote --quote quote.bin

API Endpoints

Health

GET /health - Node health status
GET /gpu/status - GPU availability and stats

Jobs

POST /jobs - Submit new job
GET /jobs/:id - Get job status
GET /jobs/:id/proof - Get 32-byte proof

Workers

POST /workers/register - Register GPU worker
GET /workers - List workers with GPU info

Deployed Contracts (Starknet Sepolia)

Core Infrastructure

Contract	Address	Purpose
StwoVerifier	`0x575968af96f814da648442daf1b8a09d43b650c06986e17b2bab7719418ddfb`	Circle STARK proof verification
ProofGatedPayment	`0x7e74d191b1cca7cac00adc03bc64eaa6236b81001f50c61d1d70ec4bfde8af0`	Payment gating for proofs
PaymentRouter	`0x001a7c5974eaa8a4d8c145765e507f73d56ee1d05419cbcffcae79ed3cd50f4d`	Fee distribution (80/18/2)
ProverStaking	`0x3287a0af5ab2d74fbf968204ce2291adde008d645d42bc363cb741ebfa941b`	Worker stake management

Token & Oracle

Contract	Address	Purpose
SAGE Token	`0x072349097c8a802e7f66dc96b95aca84e4d78ddad22014904076c76293a99850`	Native token
OracleWrapper	`0x4d86bb472cb462a45d68a705a798b5e419359a5758d84b24af4bbe5441b6e5a`	Price feeds (w/ fallback)
Faucet	`0x62d3231450645503345e2e022b60a96aceff73898d26668f3389547a61471d3`	Testnet token faucet

External Tokens (OTC Config)

Token	Address	Source
USDC	`0x053b40a647cedfca6ca84f542a0fe36736031905a9639a7f19a3c1e66bfd5080`	Bridged
STRK	`0x04718f5a0fc34cc1af16a1cdee98ffb20c31f5cd61d6ab07201858f4287c938d`	Native
WBTC	`0x00452bd5c0512a61df7c7be8cfea5e4f893cb40e126bdc40aee6054db955129e`	StarkGate

Fee Distribution Model

Payment Flow (pay_with_sage):
├── Worker: 80% → Direct SAGE transfer
├── Treasury: 18% → Protocol development
└── Stakers: 2% → Staking rewards pool

Related Resources

Validator Deployment Guide - Detailed deployment guide
GPU Operator FAQ - Common questions
Staking Guide - Token staking details
BitSage Website - Main website
Discord - Community support

License

MIT License - see LICENSE for details.

Built by BitSage Network

Powering verifiable computation with GPU-accelerated ZK proofs

1,237 proofs/sec on 4x H100 | 107M proofs/day

Website | Discord | Twitter

Name		Name	Last commit message	Last commit date
benches		benches
config		config
data/consensus		data/consensus
deploy/aws		deploy/aws
docs		docs
examples		examples
logs		logs
migrations		migrations
scripts		scripts
src		src
templates		templates
tests		tests
.env.example		.env.example
.env.mainnet.example		.env.mainnet.example
.gitignore		.gitignore
CONSENSUS_IMPLEMENTATION_COMPLETE.md		CONSENSUS_IMPLEMENTATION_COMPLETE.md
CONSENSUS_INTEGRATION_STATUS.md		CONSENSUS_INTEGRATION_STATUS.md
Cargo.toml		Cargo.toml
Dockerfile.coordinator		Dockerfile.coordinator
Dockerfile.worker		Dockerfile.worker
Dockerfile.worker.gpu		Dockerfile.worker.gpu
E2E_PRODUCTION_PLAN.md		E2E_PRODUCTION_PLAN.md
EVENT_INDEXER_AUDIT.md		EVENT_INDEXER_AUDIT.md
GPU_PROOF_PIPELINE.md		GPU_PROOF_PIPELINE.md
HIGH_PRIORITY_ITEMS_COMPLETE.md		HIGH_PRIORITY_ITEMS_COMPLETE.md
INTEGRATION_COMPLETE.md		INTEGRATION_COMPLETE.md
MAINNET_DEPLOYMENT_CHECKLIST.md		MAINNET_DEPLOYMENT_CHECKLIST.md
OBELYSK_PROOF_PACKAGE.json		OBELYSK_PROOF_PACKAGE.json
OBELYSK_STWO_PROOF_DEMO.md		OBELYSK_STWO_PROOF_DEMO.md
PHASE1B_COMPLETE.md		PHASE1B_COMPLETE.md
PHASE1_COMPLETE.md		PHASE1_COMPLETE.md
PHASE1_FINAL_SUMMARY.md		PHASE1_FINAL_SUMMARY.md
PHASE1_PROGRESS.md		PHASE1_PROGRESS.md
PRODUCTION_IMPROVEMENTS.md		PRODUCTION_IMPROVEMENTS.md
PROMETHEUS_METRICS.md		PROMETHEUS_METRICS.md
PROOF_SUBMISSION_RESULTS.md		PROOF_SUBMISSION_RESULTS.md
README.md		README.md
STARKNET_PROOF_DEMO.json		STARKNET_PROOF_DEMO.json
STARKNET_PROOF_PACKAGE.md		STARKNET_PROOF_PACKAGE.md
TLS_SETUP.md		TLS_SETUP.md
UI_BACKEND_INTEGRATION_GAPS.md		UI_BACKEND_INTEGRATION_GAPS.md
add_pair_db.sql		add_pair_db.sql
alerts.yml		alerts.yml
benchmark_results.json		benchmark_results.json
demo_p2p.sh		demo_p2p.sh
demo_script.sh		demo_script.sh
prometheus.yml		prometheus.yml
proof_package_for_starknet.json		proof_package_for_starknet.json
rust-toolchain.toml		rust-toolchain.toml
test_coordinator.sh		test_coordinator.sh

Uh oh!

Latest commit

History

Folders and files

Repository files navigation

BitSage Network - GPU Worker Node

One-Command Deployment

Option 1: Docker (Recommended)

Option 2: Native Installation

Manual Installation

Requirements

Build from Source

What the Setup Wizard Does

Start Earning

Staking Tiers

Performance Benchmarks

Latest Results (February 3, 2026)

Proof Generation Performance

On-Chain Verification Stats

Verified Transactions (Live on Starknet Sepolia)

Multi-GPU Scaling (4x H100)

GPU Comparison

Architecture

How It Works

Obelysk Proof Pipeline

Obelysk Protocol: Understanding the Proving System

Overview

Why Two Layers?

Developer Guide: Generating Proofs

Quick Start: Generate Your First Proof

CLI Proof Generation

Rust API: Programmatic Proof Generation

GPU-Accelerated Proving

Submitting Proofs to Starknet

Available Workload Types

Module Reference

Core Obelysk Modules

GPU Modules

Starknet Modules

Privacy Modules

CLI Reference

sage-worker Commands

bitsage-proof Commands

Configuration

Worker Configuration (~/.bitsage/worker.toml)

Environment Variables

Building from Source

Standard Build (CPU only)

GPU Acceleration (NVIDIA CUDA)

Multi-GPU Support

All Features

Feature Flags

Docker Deployment

Using Docker

Docker Compose

Cloud Deployment

AWS (EC2 P4d/P5)

Lambda Labs

Brev.dev (Kubernetes)

Monitoring

Built-in Dashboard

Prometheus Metrics

Grafana Integration

Troubleshooting

Common Issues

Logs

Security

Key Management

TEE Attestation

API Endpoints

Health

Jobs

Workers

Deployed Contracts (Starknet Sepolia)

Core Infrastructure

Token & Oracle

External Tokens (OTC Config)

Fee Distribution Model

Related Resources

License